OPEN: Intrafraction Motion with Open vs. Closed H&N Face Masks

Ciaran Malone (00:03):

I’m delighted to share our experience in using SRT for our open face mask trial, which ran in St. Luke’s Radiation Oncology Network in Dublin. So these are my disclosures. Just a little bit about our network. So we’re three site network treating roughly 5,500 patients per year, six CT scanners and 14 linear accelerators. So, for us to move and shift our practice towards a whole new way of treating our head and neck cases, it was important for us to do it in an evidence-based way. So for us, you know, a clinical trial doesn’t always have to be dose or toxicity related. It can answer a simple question: can we treat patients more robustly or as robustly, and also improve the quality of care to our patients too? So that’s a question we try to answer with this trial. Something physicists might lean might not quite start with is the patient.

Ciaran Malone (00:55):

So that’s really important, and we didn’t have to go too far to see that. Head and neck patients do struggle a lot with these close-fitting face masks day to day in treatment. And I mean, a lot of the radiation therapists here will, I have to go too far to find quotes. I mean, here, having the mask on didn’t worry me until they clamped my head down, and I wouldn’t move. And then from an OT perspective, this is one note that we, that you had to look for a few minutes for and find, which is that we tried unsuccessfully to treat a patient. We gave advice on relaxing and breathing techniques. The patient had traumatic childhood experiences, was very afraid of being in a mask, had panic attacks on three occasions with mask had to be removed at speed.

Ciaran Malone (01:36):

So for us, moving to open face masks was incredibly important if we could do it robustly. And it was great to see Lisa’s talk and crystal’s talk yesterday, kind of emphasizing the same points, which is really, really good. Where we started was that we wanted to have a quick move. So we looked at just moving to a simple open face mask, still maintaining the shoulder mobilization. And Dr Ashing Glynn led this study where we just looked at the setup motion, but we didn’t want to just stop here. This was a solution for the most claustrophobic patients in our network. We wanted to make a full change in practice for all head and neck patients in our network. So what this didn’t do was look at the intra-fraction motion and also really robustly measure the patient experience and distress levels do they actively actually decrease.

Ciaran Malone (02:19):

So, to do it, we didn’t want to just take one step towards an open face mask. We wanted to take the next step as well to get to a fully open three-point face mask where the shoulders were not immobilized. So with SGRT for this, we did want to reduce the complexity of the treatment setup, have something to help us get those shoulders into the right position, maybe using that deformation workflow. We wanted to track the patient’s motion where it mattered most during treatment. And we also wanted to make a little bit of a move towards a three mil margin instead of a four. So we wanted to do this in as robustly way as possible to try and put evidence behind reducing our margins for all radical head and neck patients across our network. Sam one, a very leading radiation therapist in this project, and she ran a pilot study and the real, the one thing she really wanted to see was what the optimal ROI approach for us in our network.

Ciaran Malone (03:06):

She wanted to inform staff training, and the solution she came up with was a two ROI approach. So one, just to set up the, the phase first and then to move towards monitoring for treatment ROI, which is a composite ROI, the setup for us was a two-part setup. So we got the patient’s neck inflexion correct with the facial ROI first, and then we moved across to a composite ROI then just to monitor for treatment but also just make sure everything was still in place. And the composite ROI as well was important for us to get those nodal regions in as well because many of those head neck cases do have nodal regions that come right down. We did offer guardrails, I wouldn’t say strict guidance for ROI design and selection, because I do think it’s really important for the radio radiographer and radiation therapist to have some leeway to make sure that these ROIs are bespoke for a patient and having very prescriptive rules might prevent them from doing so.

Ciaran Malone (03:56):

So we had kind of guardrails just to kind of give them some rough advice that for each patient, we make sure that the ROI is suitable and is giving us the information that we need. Then the next problem was what was a good ROI? So when you go to head and neck face masks, there are loads of different brands, there’s loads of different openings for the facial region. And we went with Orfit, but we wanted this study to be translatable across different masks and different institutions too. So the question then came up was, well, what is good topography inside these small regions of interest? Do we still maintain enough good topography to track these patients treat their treatment? And you can see that’s a registration-type problem. So you’re registering a reference surface with a surface capture, and you’re just trying to make sure that any motion between those two registrations is detected.

Ciaran Malone (04:41):

So for us, a flat ROI might slide across and not detect any motion, but one that has a lot of topography would. So if you imagine, say, a pendulum breast versus a post mastectomy breast, Sam, our lead radiation therapist on this, it was, it could have emphasized the value of this type of meeting because it was actually here where we were discussing this exact study and trying to define what good topography was. So poor Sam had to go on a sabbatical to the Norwegian mountains where she, you know, had a very arduous task of pondering what this means. And she came back with the idea of geographical topography. So we already have a solution for what good topography is or at least quantifying it, and it’s in the geographical realm. So we are looking to see whether we could we apply those ideas to the ROIs that we used in our open face mask to see if the ruggedness, the slope aspects all change much when you shorten that ROI or did they stay quite, you know, robust and stable.

Ciaran Malone (05:38):

That led to this study, which is really, it was a bit of great bit of fun from a physics perspective because I love translating different ideas across different domains. But it allowed us to quantify topography using slope aspect, ruggedness metrics. We could quantify ranges of, of topography within each different clinical site. But importantly for the head and neck cases, the more we shrank that ROI down, as long as we got the nose and the bridges, the cheeks in the topography didn’t really change much, which is great, at least from a quantitative point of view. So now the next bit of work for us is to validate this across known motions and see, you know, did that also maintains for accurately measuring or detecting motion. So this is Sam. She greatly staged the, the set of all three masks for us. This was our five-point open mask, our three-point open mask and our five-point open mask.

Ciaran Malone (06:31):

The trial design was over two years for 230 patients. We randomized equally across three arms. We did exclude patients with non claustrophobia. So the reason we did that was, and it might be a little bit counterintuitive because this is the patient cohort that would benefit most from this technique, that if we randomized the patient into the closed face mask arm, they’re very likely to drop out. So it would leave us with a different paper patient population in each arm. So that was important to us to at least if we’re comparing like for like we wanted a similar patient population in each arm that would report similarly or at least be comparable. And the beauty of it as well for us was that it was very data-centred. So we could pull out data on a per fraction basis for each patient, which allowed us to have an automated data paring pipeline.

Ciaran Malone (07:16):

So that allowed us then to track these patients’ motion continuously and see whether we need to drop an arm or action if these patients were not setting up correctly, if there’s anything that would, you know, be highlighted as a problem. So our primary objective was to compare the accuracy and our secondary objectives. Then, looking at the patient experience, comfort, tolerability, and distress. Now we originally had those flipped, and it was really important to have patient involvement in our trial because the whole reason they flipped was the patient’s ex-patient’s advice. And input the patient immediately after giving the pill. Had the patient information leaflet and the trial details said I wouldn’t sign up for this trial. And we were wondering why, why is that the case? And it was because he said I wouldn’t trade comfort for a set of accuracy. So that’s what was important for our patients to start with the set of accuracy first to assure them that we were going to give them the most robust and accurate treatment first and foremost and then measure your to tolerance and comfort and overall experience.

Ciaran Malone (08:13):

So we flipped it and that’s what was important. So we, we could assure them that we were tracking them using a synergistic approach, using SDR t com, VM CT for pre and post and reassured them that we had a lot of interim safety points as well. So it was really, really good to have a patient involvement because it might change how you actually tackle a lot of these trials or different projects. The next problem that came up was inter fraction motion. And from a physics perspective, at least for me, I was like, well that’s just the motion that happens during treatment. What’s the problem? This is going to be really easy to measure. And when we dug into the literature and kind of said “How are we going to compare our results to others?” It turns out to be a little bit of a rabbit hole we went in because how interfraction motion is reported is very variable, depending on the paper you look, look at or who measures it.

Ciaran Malone (08:57):

So you have pre and post-imaging, but that doesn’t give you motion when it matters most during the beam-on time. Then you have portal syn imaging or syn imaging, you know, at different time points. Again, you’re just hitting time points, but then how that’s wrapped up in a metric or a measure is very important. And we also have continuous motion like for SGRT and again, how do you report that is very variable across different papers. Do you report max motion or mean motion? If you report mean motion, the patient could equally vary both sides of zero, and it reports zero. So for us it was important to take a very overall approach and just report everything we could to, to fully quantify it so that any paper in the future could compare their results to ours. And this is why we took a synergistic approach, meaning for pre and post cone beam CT patient hops up on the bed, you get your first cone beam ct, the patient hops off the bed, goes to the toilet, comes back, has a tot with someone outside, hops back up again and the cone beam isn’t on the wiser.

Ciaran Malone (09:51):

So we were not happy with just using pre and post-imaging. It had to be SJT in the middle two. So that’s why we took, as I said, a synergistic approach. And our interim safety point was about 50 patients in, and we just wanted to see if we could drop the pre the post CBCT and what we could learn from the evaluation of both.

Ciaran Malone (10:13):

So, for SGRT what we found was that we could measure using the data, we could export from the SGRT system, the setup time, where you can see we’re moving up and down the bed and getting the patient into the right position. And then just this treatment portion as well where if you zoom into it and just flip it over, you can see the little green portions if it comes across well on the screen. And that was when the patient took a big deep breath into breath hold and they could hold their breath. This is a, a breath case just to kind of give you an idea of what we could measure. So that green portion is what we wanted to take out during BM on how much did this patient moved. That’s what was important to us. So you could see we could take two different angles to our perspectives on it so we could do a drift over time.

Ciaran Malone (10:48):

If a patient drifted during the course of treatment, we could measure that and we could also measure the distribution of motion as well. So that was quite interesting too. So again, we got a lot of rich data at a fraction level then we could build it up per fraction and give us a really nice, rich amount of data on a how a patient varies over the course of treatment? And we could see then, you know, did the patient progressively get worse or was there a time point of maybe 15 days in where they’re losing weight and getting a bit more unstable in a mask. And then we could look at on a population basis, so zoom out again so we could get fraction patient and population rich data out to really show us how and how much these patients moved. And then as well how a population margin might fit into this as well. Is a population margin suitable for all patients? So we could report it in many different ways. Toppers are co CT just traditionally reported as a distribution of, of, of differences pre and post and below. I just show 95th percentile as well just to give you an idea. But what SGT showed us was transient deviations that was missed by pre and post-imaging and SGRT also picked up on rotational differences that the com CTs did not pick up on as well.

Ciaran Malone (12:00):

So it allowed us to move from quotes like this, where the mask was tight, couldn’t breathe, just really struggled, to a place where I got the one with the open face mask. It’s brilliant. It was very easy and I’m amazed at how easy it was. So that was really, really nice to come ahead of the trial now, which is closed or due to be written very, very soon. But it was really nice to be able to see not only the impact from a data point of view, but also the patient point of view, too. And finally, just wanted to kind of highlight just, you know, if you get rich data to this level, I think is really useful to use it and to see what we, where we can make the next practice change. And hopefully this is what we’re looking at now, you know, we’ve all this data on a fraction level, patient level and treatment level.

Ciaran Malone (12:40):

And when you look at all the different individual PTV margins, if you’re to calculate them on a patient basis, there are some patients the PTV margins are too small, and some patients the PT v PTs are too big for. So I do think there’s evidence now that we do have the technological capabilities to move towards more of an individualized margin for these head and neck patients. So this is the next move for using our SRT system is to try and predict it and we’re using a Bayesian neural network and the idea of that is not just one answer out, you’re not just predicting this patient should get a two mil margin, it comes out and it builds up data over time. So it allows you to use the SRT data that builds up over the course of a patient’s treatment, not only to show, you know, would a patient need a bigger margin or smaller margin, but how confident we are in that.

Ciaran Malone (13:22):

So it could predict a two mil margin with a very narrow band and we would like, we have good confidence that the model is very sure that we’re in a good place to reduce this patient’s margin or a big wide confidence band, in which case we’re not sure given the traditional and make sure we monitor them a bit closer. So thank you so much for your time. I think we’ve, a number of different nice outcomes from this SRT allowed us to move into a three point open face mask place, you know, confidently, it did quantify motion when it matters most during BM on it did allow us now to look at personalized margins and that’s the next move for us in a, in a clinical trial setting. And it also allowed us not only to look at the motion but the surfaces were measuring as surrogates and show that, you know, even small facial or eyes do give good surface topography. So thank you so much for your time. We really appreciate it and I’d be happy to answer any questions.